I build AI into the way your business already works: assistants that answer from your own data, automations that do the routine work, and guards on anything that changes something.
I already run this on production servers, with several guards between a prompt and a live change.
Before · 2–4 hours
# Update the opening hours on three client sites
ssh server
cd /var/www/client-one
nano footer.php
# ...repeat for each site, then check by handNow · minutes
# The same update, from a prompt
assistant.run("Update the opening hours")
guards: "checked"
approval: "required"
status: "live"AI usually pays off first in one of these three places. Pick the line that sounds like you. Prices are starting points, and every project gets a fixed quote once I know the scope.
From $500
Fixed quote
An assistant on your website, WhatsApp, or inbox that answers from your own products, prices, and policies, and hands the conversation to a person when it should.
From $400
Fixed quote
AI steps inside the processes you already run: reading requests, drafting replies, updating records, and making routine changes to your website, with a check before anything goes live.
From $600
Fixed quote
Search and answers over your own documents, procedures, and product catalogs, so your team finds what it needs in seconds and every answer points to its source.
In most businesses the slow part is work that people still do by hand.
5 warnings found
AI that changes things needs limits. Everything I build checks its work before it acts, asks a person when it should, and leaves a record you can read.
I use standard model APIs, Hermes, OpenClaw, or n8n when they fit, and build from scratch when your business needs functions they don't have.
Four steps, and you know the price before any work starts.
There's no account manager in between, and nothing gets handed to a junior.
I currently work remotely on tracking infrastructure for a marketing agency in Hamburg, and I've built and maintained more than 75 websites since 2020.
I work with AI tools every day, and the same guarded setup already runs on production servers. It keeps turnaround short.
I'm based in Yogyakarta (UTC+7), so my afternoons overlap with mornings in Europe.
My afternoon, 12:00–18:00 in Yogyakarta, is the morning in Berlin.
Your data is used to answer and act for your business, not to train public models. I choose providers and settings for that, and document where your data goes.
Standard model APIs, Hermes, OpenClaw, or n8n when they fit, and a custom build when your business needs functions they don't have.
Guards check changes before they go live, a person approves anything outside the agreed limits, and every action is logged so it can be traced and undone.
Yes. I work with agencies on their clients' projects, under your name if you prefer, delivered to your standard and documentation.
Every service has a starting price, and every project gets a fixed quote after a free intro. Prices are in US dollars, and can be switched to euros or Indonesian rupiah. Visitors in Indonesia see rupiah first. Running costs, such as model usage, are explained before anything is built.
Contact
Tell me what you have and where the hours go. I usually reply within one working day with how I'd approach it and a quote.
Prefer email? Write to hello@sultankautsar.com.